SPAN and Futures Margin Models: Scanning Ranges, Spread Charges, and SPAN 2

SPAN, short for Standard Portfolio Analysis of Risk, is the model CME Group’s clearing house has used since 1988 to set futures margin: it revalues every position in a portfolio under 16 hypothetical market scenarios, takes the worst loss, and then adds charges and subtracts credits for risks the scenarios miss or overstate. SPAN 2, its successor, replaces most of that fixed grid with a filtered historical value-at-risk model, and as of the end of 2025 CME had moved its energy and equity products onto it, with interest rates, FX, agriculture and the remaining commodities still to come.

In short: the margin on a futures position is not a fixed percentage of its value. A clearing house estimates how much the position could lose over the time it would need to close it out if the owner defaulted, to at least 99% confidence, and requires that amount as a deposit. Under SPAN the estimate is the largest loss across 16 price and volatility scenarios, plus a charge for calendar spread risk, plus a delivery-month charge, minus credits for offsetting positions in related products, floored by a minimum for short options. Under SPAN 2 the estimate comes from rescaled historical returns, stress scenarios, and explicit charges for liquidity and concentration. Either way, the number moves with volatility and with the price level, and the clearing house can change it overnight.

This builds on how futures margin works (initial and maintenance margin, daily settlement, margin calls). Here we open the model, run a full SPAN calculation, and look at what SPAN 2 changes.

A common misconception: “margin is the exchange’s percentage of contract value”

Retail platforms show one number per contract, which makes margin look like a down payment set as a share of notional value. The clearing house actually sets requirements from the risk of the whole portfolio. Two traders holding the same March corn contract can face very different requirements, because a long March and a short July corn position mostly cancel out, and the model knows that.

Margin is also not a cost. It is collateral covering the clearing house’s exposure between your last variation payment and the moment it could liquidate you. Losses beyond that are still yours (see futures risks explained honestly).

Who sets the number and to what standard

CME Clearing is registered with the Commodity Futures Trading Commission as a derivatives clearing organization, and as a systemically important one it is subject to enhanced standards. In a December 2023 presentation to the CFTC’s Agricultural Advisory Committee, CME stated the core standards its models are built to:

Performance bond requirements are designed to cover 99% of market moves on an ex post basis over a defined period of risk.
The margin period of risk for exchange-traded futures and options is at least one day.
Requirements are recalculated at least once a day for all products, and twice daily for most.
CME’s portfolio coverage for the twelve months ending in the third quarter of 2023 was 99.97% for its base futures and options products.

Those standards come from CFTC Regulation 39.13(g): initial margin commensurate with each product’s risk, a minimum one-day liquidation time for futures, and at least 99% coverage verified by backtesting.

The clearing house collects margin from its clearing members, the futures commission merchants that carry customer accounts, which collect from customers and can charge more. At CME, the customer initial requirement equals the maintenance level for accounts without a “heightened risk profile” and is 110% of maintenance for accounts with one; the clearing firm decides which category an account falls into. Brokers layer house requirements on top, especially around reports, holidays and delivery periods. Where the clearing house fits in the default waterfall is covered in how clearing houses work.

How classic SPAN works

CME’s own SPAN methodology presentation (a 2010 CME Group deck, still the clearest public description) lays out the structure. SPAN groups every instrument on the same underlying, futures and options on futures alike, into a combined commodity. It then works in two steps: first, it measures the risk of each combined commodity on its own; second, it looks for risk-reducing offsets between combined commodities.

The 16-scenario risk array

For each contract, SPAN computes a risk array: the gain or loss under 16 scenarios over a one-day look-ahead. The exchange sets two parameters per product:

the price scan range, an estimate of a large one-day move in the underlying, stated in dollars per contract, and
the volatility scan range, how far implied volatility might move, which matters for options.

The 16 scenarios test the underlying unchanged and moved up and down by one third, two thirds and the full price scan range, each paired with volatility up and down (14 scenarios). The last two are extreme scenarios: CME moves the price by three times the scan range and counts only 33% of the resulting loss. They exist to catch deep out-of-the-money short options, which can show almost no risk inside the normal range and then explode when a market gaps.

The worst loss across the 16 rows is the scan risk for that combined commodity.

What the scan misses: the add-ons and credits

The scan assumes all months of a product move together perfectly. They do not, so SPAN adds:

Intra-commodity (inter-month) spread charge. SPAN counts how many calendar spreads the portfolio implicitly contains, using each month’s net delta, and charges a fixed amount per spread. Charges can be tiered by month groups or set for specific pairs, so a March/September spread can cost more than a March/December one. CME’s deck shows the logic with a Eurodollar calendar spread: the two legs cancel in every scenario, scan risk is zero, and the $200 spread charge becomes the whole requirement.
Delivery (spot month) charge. An add-on for contracts in or near delivery, when volatility tends to rise and the legs of a spread decouple.
Inter-commodity spread credit. Credit for offsetting positions in related products. CME’s example: long one S&P 500 future against short two Nasdaq futures, with a recognised spread ratio of 1:2 and an 85% credit, cuts a $50,500 outright requirement to $7,575.
Short option minimum (SOM). A floor per short option, so a far out-of-the-money short option is never free.

The requirement for each combined commodity is the greater of (scan risk + intra-commodity charge + delivery charge - inter-commodity credit) and the short option minimum. The portfolio total is the sum across combined commodities.

Net option value

Option premiums are handled through net option value (NOV), long option value minus short option value. Long options reduce the requirement; short options increase it, because the account must hold enough to buy them back and still cover the scenario loss.

Worked example: a full SPAN calculation on a corn portfolio

Corn is a good test case because, as of CME’s most recent annual report, agricultural products were still margined under classic SPAN. Every parameter below is illustrative: CME publishes the real ones in its daily SPAN risk parameter files, and they change. The contract specs are real: CBOT corn is 5,000 bushels, so one cent per bushel is $50 per contract.

Portfolio:

Long 10 March 2027 corn futures at $4.50 per bushel
Short 6 July 2027 corn futures at $4.65
Short 4 March 2027 corn 480 calls (strike $4.80)

Illustrative SPAN parameters:

Parameter	Value
Price scan range	30 cents per bushel ($1,500 per contract)
Volatility scan range	5 volatility points
Implied volatility of the 480 call	25%
Days to option expiry	101 (to February 19, 2027)
Intra-commodity spread charge	$250 per March/July spread
Short option minimum	$50 per short option
Extreme scenarios	3 x scan range, 33% of the loss counted

The call is priced with the Black-76 model for options on futures, with a 4% discount rate. Its value is 12.12 cents, or $605.81 per option, and its delta is about 0.33.

Step 1: the risk array

Each scenario moves both corn months by the same amount (that is the scan’s simplifying assumption) and revalues the calls one day later at the shocked volatility. Gains are positive, losses negative, in dollars:

#	Price move	Volatility	Futures (10 Mar long, 6 Jul short)	4 short calls	Portfolio
1	Unchanged	Up	0	-844	-844
2	Unchanged	Down	0	+844	+844
3	Up 1/3	Up	+2,000	-1,629	+371
4	Up 1/3	Down	+2,000	+204	+2,204
5	Down 1/3	Up	-2,000	-165	-2,165
6	Down 1/3	Down	-2,000	+1,342	-658
7	Up 2/3	Up	+4,000	-2,521	+1,479
8	Up 2/3	Down	+4,000	-589	+3,411
9	Down 2/3	Up	-4,000	+412	-3,588
10	Down 2/3	Down	-4,000	+1,713	-2,287
11	Up full	Up	+6,000	-3,519	+2,481
12	Up full	Down	+6,000	-1,541	+4,459
13	Down full	Up	-6,000	+893	-5,107
14	Down full	Down	-6,000	+1,978	-4,022
15	Up 3x (33%)	Unchanged	+5,940	-3,553	+2,387
16	Down 3x (33%)	Unchanged	-5,940	+782	-5,158

The futures column is simple: the portfolio is net long four contracts, and each full scan move is 30 cents, or $1,500 a contract. The options column is not linear, which is the point of a scenario grid. The worst row is the extreme down scenario, a 90-cent fall counted at 33%, so scan risk is $5,158. Scenario 13 is a close second at $5,107.

Step 2: the calendar spread charge

The scan treated the March long and July short as perfectly offsetting. To measure how many spreads that assumption covers, SPAN uses composite deltas: each contract’s delta is averaged across the price scan points with fixed weights. CME’s deck gives its weights as 0.27 for unchanged, 0.217 for each one-third move, 0.11 for each two-thirds move and 0.037 for each full move.

The composite delta of each 480 call comes to 0.333.
March net delta: 10 - (4 x 0.333) = 8.67 long.
July net delta: 6 short.
Spreads formed: the smaller of the two, 6.
Charge: 6 x $250 = $1,500.

Step 3: the floor, the total, and option value

Scan risk plus spread charge: $5,158 + $1,500 = $6,658. No inter-commodity credit applies, and no delivery charge, since March is not yet in delivery.
Short option minimum: 4 x $50 = $200, well below, so it does not bind. SPAN risk is $6,658.
Net option value: 4 short calls x $605.81 = -$2,423.
Maintenance requirement: $6,658 + $2,423 = $9,081.
For an account classed as heightened risk, initial margin at 110%: $9,989.

The NOV term is mostly the trader’s own money: selling the four calls put $2,423 of premium into the account, and the requirement keeps it there to buy them back.

What the model saved, and what a margin change does

Margining each of the 16 futures legs as an outright position at $1,500 apiece would require $24,000 before the options. SPAN charges $9,081 for the whole portfolio, because the July short offsets most of the March long and the short calls offset some of the remaining long delta. A pure 10-lot March/July spread with no options would cost just the spread charge, 10 x $250 = $2,500, against $15,000 for ten outright contracts.

Now suppose volatility rises and CME widens the corn scan range by 20%, to 36 cents. Rerunning the same grid gives scan risk of $6,334, SPAN risk of $7,834, and a maintenance requirement of $10,257, up 13%. The increase is smaller than 20% because the spread charge did not change. A separate increase in the spread charge would hit spread traders directly, which is exactly what happens when calendar spreads, rather than outright prices, become volatile.

How CME calibrates the parameters

CME’s 2023 CFTC presentation describes the calibration for classic SPAN in some detail:

Parameters are set product by product from weighted lookback periods, with observations over one month, three months, six months, one year, four years, ten years or longer, depending on the product.
A one-year lookback, for example, takes the day-over-day settlement price changes in dollars and sets margin to cover the move at at least 99% confidence.
To damp procyclical swings, it uses seasonality factors (such as new-crop and old-crop dynamics in grains), volatility floors, buffers above the baseline, multiple lookbacks, and implied volatility from options.
CME also accounts for event risk that historical data may not show, and, when futures settle at a price limit, may use synthetic prices derived from options. (How limits interact with margin is covered in limit moves and circuit breakers.)

The same presentation makes a point retail traders often miss: holding volatility fixed, margins scale with notional value. If the price of corn doubles, the dollar margin per contract doubles too, all else equal.

CME announces new performance bond levels in clearing advisory notices before they take effect; brokers then apply them, often with their own add-on.

SPAN 2: what changes

CME’s 2024 certification to the CFTC applying SPAN 2 to equity products (CME Submission No. 24-425, filed October 2, 2024) is the most detailed public description of the new framework. Its central change is the market risk model: SPAN 2 uses filtered historical value-at-risk (FHVaR). It takes actual historical returns, divides each by the volatility prevailing when it happened, and multiplies by today’s volatility, using an exponentially weighted moving average (EWMA) volatility estimate. In CME’s words, in a highly volatile period the framework scales up returns observed in calmer periods, and does the opposite when markets are quiet.

Illustration: how filtering rescales history

Assume current EWMA volatility is 28% annualised. Five historical daily returns, each recorded at a different volatility:

Historical return	Volatility at the time	Rescaled to today’s 28%
-3.5%	22%	-4.45%
+2.8%	18%	+4.36%
-5.2%	40%	-3.64%
+4.1%	30%	+3.83%
-2.1%	15%	-3.92%

The 5.2% crash day from a 40% volatility regime becomes a smaller loss today, while a modest 2.1% drop from a calm 15% period becomes a large one. The model then reads a percentile off the rescaled distribution, applied to the actual portfolio rather than a fixed grid. These numbers are invented; the filing does not publish CME’s EWMA decay factor.

The three components

The 24-425 filing describes SPAN 2 as three components:

Market risk: the FHVaR, plus a stressed VaR sub-component that keeps historical scenarios from stressed periods outside the lookback window and adds hypothetical “extreme but plausible” scenarios. For equity options it keeps a short option minimum and adds a valuation uncertainty margin for gaps between model and settlement prices.
Liquidity risk: the expected cost of liquidating the portfolio, built from open interest, bid-ask spreads, volume and market surveys.
Concentration risk: the extra cost of liquidating a large position.

Product-specific settings

The same filing shows how the framework is tuned by product group:

Setting	Energy	Equity
FHVaR lookback	10 years	At least 5 years (10 years kept in stressed VaR)
Confidence level (ex ante)	99%	At least 99.5%
Correlation scaling	Used for certain products	Not used
Principal component analysis	Used	Not used (activity is concentrated in the first few months)
Samuelson adjustment	Used where the maturity effect is strong	Not used (cash settled)
Volatility floors	Yes	Yes

The Samuelson adjustment is worth a note for commodity traders. CME’s filing describes the Samuelson effect as contracts becoming more volatile as expiry approaches, which weakens the correlation between the front and second months, and notes the effect is most pronounced in physically delivered contracts. That is the same risk the classic SPAN delivery charge was built for, now modelled directly (see natural gas futures in depth for a market where it matters).

Pods and cross-model offsets

SPAN 2 margins products in groups called pods; the energy migration created a crude pod and a natural gas pod, with no offsets between them. During the transition, a product still on SPAN can get offsets against a SPAN 2 product through a cross-model offset where CME judges the correlation has a conceptual basis.

SPAN 2 rollout status, as of November 2026

Asset class	Status
Energy	Migration began in July 2023, per CME’s 24-425 filing; the CFTC presentation says core energy moved during the second half of 2023
Equities	Migration began with the end-of-day cycle for trade date October 18, 2024, excluding crypto, commodity index, legacy total return, real estate and event contracts
Interest rates, FX, agriculture, metals and other commodities	Not yet migrated as of the end of 2025, per CME Group’s annual report

CME’s earlier timelines slipped: its December 2023 CFTC presentation planned equities for the second quarter of 2024, rates and FX for the second half of 2024, and agriculture and remaining products for the second half of 2025. CME Group’s 10-K for 2025, filed in February 2026, states that “as of the end of 2025, the SPAN 2 framework has been implemented for our energy and equity products, with plans to transition interest rates, FX, agriculture and commodities,” and that customers can keep using the SPAN software, which supports both methodologies, during the transition. CME’s 2026 clearing notices through October added more energy products (including natural gas pod products in February 2026) to the SPAN 2 in-scope lists and to the cross-model offset; none migrated a new asset class. Check CME’s SPAN 2 rollout page for the current schedule.

Two practical consequences follow. Under SPAN 2 a fixed portfolio’s margin can change every day, though CME’s presentation says energy margins under the two methods were similar in level. And the equity filing expected aggregate initial margin to fall by less than 1%: SPAN 2 is meant to keep the same risk appetite, not to cut margins.

What can go wrong

Margins rise when you can least afford it. Every 99% model reacts to volatility. Floors, buffers and stressed VaR soften this, but a volatility spike still raises requirements just as positions are losing money.
Spread margin assumes the spread behaves. Calendar spread charges are small because spreads are usually stable. When a spread itself becomes the volatile thing, as in natural gas winter spreads, losses can swamp a margin set for normal conditions. Amaranth Advisors’ 2006 natural gas losses came largely from spreads between winter and non-winter contract months (see famous futures blowups).
Your broker is not the clearing house. Exchange minimums are floors; clearing firms add house margin and can raise it without notice.
Model changes move margin too. A migration from SPAN to SPAN 2 or a parameter recalibration can change requirements on an unchanged portfolio.
Margin is not a loss limit. It is sized for a one-day liquidation at 99% confidence. Larger moves happen, and losses beyond margin remain the account owner’s.

The short version

Futures margin is a clearing house estimate of how much a portfolio could lose in the time needed to liquidate it, to 99% confidence or better. Classic SPAN revalues the portfolio under 16 scenarios, adds spread and delivery charges, subtracts inter-commodity credits, and floors short options. In our corn example that produced $9,081 of maintenance margin for a portfolio whose legs would cost $24,000 margined separately, and a 20% wider scan range raised it 13%. SPAN 2 replaces the grid with filtered historical VaR plus stress, liquidity and concentration components, and as of the end of 2025 CME had moved energy and equities to it, with rates, FX, agriculture and metals still on classic SPAN. Margins move with volatility and price, and protect the clearing house first. This explains a mechanism; it is not a guide to how much risk to take.